In [1]:
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import warnings
warnings.filterwarnings('ignore')
from scipy.stats import zscore

JANUARY FLIGHT DELAY PREDICATION¶

image.png

This is the first part of flight delay prediction i.e. for the month of January. To check the second part i.e. for the month of February, please have a look at this dataset: February Flight Delay Prediction¶

This data is collected from the Bureau of Transportation Statistics, Govt. of the USA. This data is open-sourced under U.S. Govt. Works. This dataset contains all the flights in the month of January 2019 and January 2020. There are more than 400,000 flights in the month of January itself throughout the United States. The features were manually chosen to do a primary time series analysis. There are several other features available on their website.¶

This data could well be used to predict the flight delay at the destination airport specifically for the month of January in upcoming years as the data is for January on¶

In [2]:
df=pd.read_csv('fl.csv')
df
Out[2]:
DAY_OF_MONTH DAY_OF_WEEK OP_UNIQUE_CARRIER OP_CARRIER_AIRLINE_ID OP_CARRIER TAIL_NUM OP_CARRIER_FL_NUM ORIGIN_AIRPORT_ID ORIGIN_AIRPORT_SEQ_ID ORIGIN ... DEST DEP_TIME DEP_DEL15 DEP_TIME_BLK ARR_TIME ARR_DEL15 CANCELLED DIVERTED DISTANCE Unnamed: 21
0 1 3 EV 20366 EV N48901 4397 13930 1393007 ORD ... GRB 1003.0 0.0 1000-1059 1117.0 0.0 0.0 0.0 174.0 NaN
1 1 3 EV 20366 EV N16976 4401 15370 1537002 TUL ... ORD 1027.0 0.0 1000-1059 1216.0 0.0 0.0 0.0 585.0 NaN
2 1 3 EV 20366 EV N12167 4404 11618 1161802 EWR ... TYS 1848.0 0.0 1800-1859 2120.0 0.0 0.0 0.0 631.0 NaN
3 1 3 EV 20366 EV N14902 4405 10781 1078105 BTR ... IAH 1846.0 0.0 1800-1859 2004.0 0.0 0.0 0.0 253.0 NaN
4 1 3 EV 20366 EV N606UX 4407 14524 1452401 RIC ... IAH 1038.0 0.0 1000-1059 1330.0 0.0 0.0 0.0 1157.0 NaN
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
607341 31 5 9E 20363 9E N331CA 4812 15412 1541205 TYS ... DTW 1002.0 1.0 0700-0759 1128.0 1.0 0.0 0.0 443.0 NaN
607342 31 5 9E 20363 9E N295PQ 4813 11433 1143302 DTW ... JFK 1747.0 0.0 1700-1759 1933.0 0.0 0.0 0.0 509.0 NaN
607343 31 5 9E 20363 9E N294PQ 4814 11996 1199603 GSP ... LGA 554.0 0.0 0600-0659 752.0 0.0 0.0 0.0 610.0 NaN
607344 31 5 9E 20363 9E N228PQ 4815 10397 1039707 ATL ... XNA 1714.0 0.0 1700-1759 1811.0 0.0 0.0 0.0 589.0 NaN
607345 31 5 9E 20363 9E N228PQ 4815 15919 1591904 XNA ... ATL 1841.0 0.0 1800-1859 2128.0 0.0 0.0 0.0 589.0 NaN

607346 rows × 22 columns

In [3]:
df.head()
Out[3]:
DAY_OF_MONTH DAY_OF_WEEK OP_UNIQUE_CARRIER OP_CARRIER_AIRLINE_ID OP_CARRIER TAIL_NUM OP_CARRIER_FL_NUM ORIGIN_AIRPORT_ID ORIGIN_AIRPORT_SEQ_ID ORIGIN ... DEST DEP_TIME DEP_DEL15 DEP_TIME_BLK ARR_TIME ARR_DEL15 CANCELLED DIVERTED DISTANCE Unnamed: 21
0 1 3 EV 20366 EV N48901 4397 13930 1393007 ORD ... GRB 1003.0 0.0 1000-1059 1117.0 0.0 0.0 0.0 174.0 NaN
1 1 3 EV 20366 EV N16976 4401 15370 1537002 TUL ... ORD 1027.0 0.0 1000-1059 1216.0 0.0 0.0 0.0 585.0 NaN
2 1 3 EV 20366 EV N12167 4404 11618 1161802 EWR ... TYS 1848.0 0.0 1800-1859 2120.0 0.0 0.0 0.0 631.0 NaN
3 1 3 EV 20366 EV N14902 4405 10781 1078105 BTR ... IAH 1846.0 0.0 1800-1859 2004.0 0.0 0.0 0.0 253.0 NaN
4 1 3 EV 20366 EV N606UX 4407 14524 1452401 RIC ... IAH 1038.0 0.0 1000-1059 1330.0 0.0 0.0 0.0 1157.0 NaN

5 rows × 22 columns

In [4]:
df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 607346 entries, 0 to 607345
Data columns (total 22 columns):
 #   Column                 Non-Null Count   Dtype  
---  ------                 --------------   -----  
 0   DAY_OF_MONTH           607346 non-null  int64  
 1   DAY_OF_WEEK            607346 non-null  int64  
 2   OP_UNIQUE_CARRIER      607346 non-null  object 
 3   OP_CARRIER_AIRLINE_ID  607346 non-null  int64  
 4   OP_CARRIER             607346 non-null  object 
 5   TAIL_NUM               606648 non-null  object 
 6   OP_CARRIER_FL_NUM      607346 non-null  int64  
 7   ORIGIN_AIRPORT_ID      607346 non-null  int64  
 8   ORIGIN_AIRPORT_SEQ_ID  607346 non-null  int64  
 9   ORIGIN                 607346 non-null  object 
 10  DEST_AIRPORT_ID        607346 non-null  int64  
 11  DEST_AIRPORT_SEQ_ID    607346 non-null  int64  
 12  DEST                   607346 non-null  object 
 13  DEP_TIME               600682 non-null  float64
 14  DEP_DEL15              600647 non-null  float64
 15  DEP_TIME_BLK           607346 non-null  object 
 16  ARR_TIME               600271 non-null  float64
 17  ARR_DEL15              599268 non-null  float64
 18  CANCELLED              607346 non-null  float64
 19  DIVERTED               607346 non-null  float64
 20  DISTANCE               607346 non-null  float64
 21  Unnamed: 21            0 non-null       float64
dtypes: float64(8), int64(8), object(6)
memory usage: 101.9+ MB
In [5]:
df.isnull().sum()/len(df)*100
Out[5]:
DAY_OF_MONTH               0.000000
DAY_OF_WEEK                0.000000
OP_UNIQUE_CARRIER          0.000000
OP_CARRIER_AIRLINE_ID      0.000000
OP_CARRIER                 0.000000
TAIL_NUM                   0.114926
OP_CARRIER_FL_NUM          0.000000
ORIGIN_AIRPORT_ID          0.000000
ORIGIN_AIRPORT_SEQ_ID      0.000000
ORIGIN                     0.000000
DEST_AIRPORT_ID            0.000000
DEST_AIRPORT_SEQ_ID        0.000000
DEST                       0.000000
DEP_TIME                   1.097233
DEP_DEL15                  1.102996
DEP_TIME_BLK               0.000000
ARR_TIME                   1.164904
ARR_DEL15                  1.330049
CANCELLED                  0.000000
DIVERTED                   0.000000
DISTANCE                   0.000000
Unnamed: 21              100.000000
dtype: float64
In [3]:
df.drop('Unnamed: 21',axis=1,inplace=True)
In [9]:
df.head()
Out[9]:
DAY_OF_MONTH DAY_OF_WEEK OP_UNIQUE_CARRIER OP_CARRIER_AIRLINE_ID OP_CARRIER TAIL_NUM OP_CARRIER_FL_NUM ORIGIN_AIRPORT_ID ORIGIN_AIRPORT_SEQ_ID ORIGIN ... DEST_AIRPORT_SEQ_ID DEST DEP_TIME DEP_DEL15 DEP_TIME_BLK ARR_TIME ARR_DEL15 CANCELLED DIVERTED DISTANCE
0 1 3 EV 20366 EV N48901 4397 13930 1393007 ORD ... 1197705 GRB 1003.0 0.0 1000-1059 1117.0 0.0 0.0 0.0 174.0
1 1 3 EV 20366 EV N16976 4401 15370 1537002 TUL ... 1393007 ORD 1027.0 0.0 1000-1059 1216.0 0.0 0.0 0.0 585.0
2 1 3 EV 20366 EV N12167 4404 11618 1161802 EWR ... 1541205 TYS 1848.0 0.0 1800-1859 2120.0 0.0 0.0 0.0 631.0
3 1 3 EV 20366 EV N14902 4405 10781 1078105 BTR ... 1226603 IAH 1846.0 0.0 1800-1859 2004.0 0.0 0.0 0.0 253.0
4 1 3 EV 20366 EV N606UX 4407 14524 1452401 RIC ... 1226603 IAH 1038.0 0.0 1000-1059 1330.0 0.0 0.0 0.0 1157.0

5 rows × 21 columns

CORRELATION BETWEEN FEATURES¶

In [14]:
plt.figure(figsize=(14,6))
corr=df.corr(method='pearson')
heatmap=sns.heatmap(corr,annot=True,vmax=1,vmin=-1,linewidths=1,linecolor='Black')
plt.show()
In [15]:
num_col=df.select_dtypes(include=("int",'float')).columns
num_col
Out[15]:
Index(['DAY_OF_WEEK', 'ORIGIN_AIRPORT_ID', 'DEST_AIRPORT_ID', 'DEP_TIME',
       'DEP_DEL15', 'ARR_TIME', 'ARR_DEL15', 'DISTANCE'],
      dtype='object')
In [16]:
df.tail()
Out[16]:
DAY_OF_WEEK ORIGIN_AIRPORT_ID DEST_AIRPORT_ID DEP_TIME DEP_DEL15 DEP_TIME_BLK ARR_TIME ARR_DEL15 DISTANCE
607341 5 15412 11433 1002.0 1.0 0700-0759 1128.0 1.0 443.0
607342 5 11433 12478 1747.0 0.0 1700-1759 1933.0 0.0 509.0
607343 5 11996 12953 554.0 0.0 0600-0659 752.0 0.0 610.0
607344 5 10397 15919 1714.0 0.0 1700-1759 1811.0 0.0 589.0
607345 5 15919 10397 1841.0 0.0 1800-1859 2128.0 0.0 589.0

Comparing Arrivals and Departures Time Frames¶

In [3]:
from plotly.subplots import make_subplots
import plotly.graph_objects as go
values = [df['DEP_DEL15'].value_counts()[0],df['DEP_DEL15'].value_counts()[1]]
labels = ["Delayed (<15 minutes)", "Delayed (>15 minutes)"]
colors = ['blue','red']
values_arr = [df['ARR_DEL15'].value_counts()[0],df['ARR_DEL15'].value_counts()[1]]

fig = make_subplots(1, 2, specs=[[{'type':'domain'}, {'type':'domain'}]], subplot_titles=['Departures', 'Arrivals'])
fig.add_trace(go.Pie(labels=labels, values=values, pull=[0, 0.1],textinfo = 'label+percent'),1,1)
fig.add_trace(go.Pie(labels=labels, values=values_arr, pull=[0, 0.1],textinfo = 'label+percent'),1,2)
fig.update_traces( textinfo='value', textfont_size=14, marker=dict(colors=colors, line=dict(color='#000000', width=2)))
fig.update_layout(uniformtext_minsize=4, uniformtext_mode='hide', title_text='Flights Delayed', title_x = 0.5, legend_title='<b>Flights</b>',legend=dict(x=.45, y=0.6))

Flight Delay On Week Days¶

In [18]:
# Set the width and height of the figure
plt.figure(figsize=(12,6))

# Add title
plt.title("Flight delay on week days")

# Bar chart showing average arrival delay for Spirit Airlines flights by month
sns.barplot(x=df.DAY_OF_WEEK, y=df['ARR_DEL15'])

# Add label for vertical axis
plt.ylabel("Arrival delay (in minutes)")
Out[18]:
Text(0, 0.5, 'Arrival delay (in minutes)')

HANDLING DATA¶

Dealing With Missing Values¶

In [ ]:
#drop flights with missing data
df = df[(df.CANCELLED != 1) & (df.DEP_TIME.isna() == False) & (df.ARR_TIME.isna() == False)]
len(df.CANCELLED.isna()), len(df.DEP_TIME.isna()), len(df.ARR_TIME.isna()), len(df)

#drop when target is NAN
df = df[(df.ARR_DEL15.isna() == False)]

df.isna().sum()
In [ ]:
df.drop('DEP_TIME',axis=1,inplace=True)
In [ ]:
df.drop('DEP_DEL15',axis=1,inplace=True)

DEP_TIME_BLK contains an invalid label fix it to 0500-0559¶

In [19]:
df.loc[df['DEP_TIME_BLK']=='0001-0559', 'DEP_TIME_BLK'] = '0500-0559'
print(df.loc[[211]])

     DAY_OF_WEEK  ORIGIN_AIRPORT_ID  DEST_AIRPORT_ID  DEP_TIME  DEP_DEL15  \
211            3              11140            12266    1207.0        0.0   

    DEP_TIME_BLK  ARR_TIME  ARR_DEL15  DISTANCE  
211    1200-1259    1259.0        0.0     201.0

Droping unnecessary colums¶

In [ ]:
cols=['OP_CARRIER_AIRLINE_ID', 'OP_CARRIER', 'DIVERTED', 'CANCELLED',
      'ORIGIN', 'TAIL_NUM', 'OP_CARRIER_FL_NUM', 'ORIGIN_AIRPORT_SEQ_ID',
      'DEST_AIRPORT_SEQ_ID', 'DEST', 'DAY_OF_MONTH', 'OP_UNIQUE_CARRIER']
df=df.drop(cols,axis=1)
In [ ]:
df.isna().sum()

Converting data into more useful form¶

In [20]:
def get_day_category(day_of_week):
    if day_of_week <= 5:
        return 0 #'for working day'
    elif day_of_week > 5:
        return 1 #'for weekend'
    
df.DAY_OF_WEEK = df.DAY_OF_WEEK.apply(get_day_category)
df.rename(columns={'DAY_OF_WEEK':'type_of_day'})
Out[20]:
type_of_day ORIGIN_AIRPORT_ID DEST_AIRPORT_ID DEP_TIME DEP_DEL15 DEP_TIME_BLK ARR_TIME ARR_DEL15 DISTANCE
0 0 13930 11977 1003.0 0.0 1000-1059 1117.0 0.0 174.0
1 0 15370 13930 1027.0 0.0 1000-1059 1216.0 0.0 585.0
2 0 11618 15412 1848.0 0.0 1800-1859 2120.0 0.0 631.0
3 0 10781 12266 1846.0 0.0 1800-1859 2004.0 0.0 253.0
4 0 14524 12266 1038.0 0.0 1000-1059 1330.0 0.0 1157.0
... ... ... ... ... ... ... ... ... ...
607341 0 15412 11433 1002.0 1.0 0700-0759 1128.0 1.0 443.0
607342 0 11433 12478 1747.0 0.0 1700-1759 1933.0 0.0 509.0
607343 0 11996 12953 554.0 0.0 0600-0659 752.0 0.0 610.0
607344 0 10397 15919 1714.0 0.0 1700-1759 1811.0 0.0 589.0
607345 0 15919 10397 1841.0 0.0 1800-1859 2128.0 0.0 589.0

599268 rows × 9 columns

In [21]:
df['origin_day_avg'] = df.groupby(['ORIGIN_AIRPORT_ID','DAY_OF_WEEK'])['DEP_TIME'].transform('mean').astype(np.float16)

df['dest_day_avg'] = df.groupby(['DEST_AIRPORT_ID','DAY_OF_WEEK'])['ARR_TIME'].transform('mean').astype(np.float16)

Replace Missing Target Values¶

In [22]:
#replace missing target values 
def new_arrival_delay(dep_delay, arr_delay):
    if np.isnan(arr_delay):
        return dep_delay
    else:
        return arr_delay

df['ARR_DEL15'] = df.apply(lambda row :new_arrival_delay(row['DEP_DEL15'],row['ARR_DEL15']), axis = 1)

CREATE ARR_TIME_BLK¶

In [23]:
# CREATE ARR_TIME_BLK 

#generate block hours
blocks = []
for hour in range(0,24):
    hour_part = ('%02d' %(hour))
    blocks.append(hour_part + '00-' + hour_part + '59')
blocks

def arrival_time_blk(arr_time):
    arr_hour = str('%04d' %(arr_time))[:2]
    arr_block = None
    for block in blocks:
        #print (block,arr_hour)
        if block.startswith(arr_hour):
            arr_block = block
            break
    if arr_block == None and str(arr_time) == '2400.0':
        arr_block = '0000-0059'
        #print('Cannot find block for #' + str(arr_time) + '#: set block to #' + arr_block + '#')
    return arr_block

df['ARR_TIME_BLK'] = df.ARR_TIME.apply(arrival_time_blk)

# drop the no more useful ARR_TIME
df.drop(['ARR_TIME'], inplace=True, axis=1)

Label Encoding Of Categorical Data Colums¶

In [24]:
from sklearn.preprocessing import LabelEncoder
le = LabelEncoder()
le.fit(blocks)
le.classes_
df['ARR_TIME_BLK'] = le.transform(df.ARR_TIME_BLK.values)
df['DEP_TIME_BLK'] = le.transform(df.DEP_TIME_BLK.values)
In [26]:
df=df.reindex(sorted(df.columns), axis=1)
In [36]:
df=df.drop(['DEP_TIME','DEP_DEL15'], axis=1,inplace=True)

Spliting Data into features and target¶

In [39]:
from sklearn.model_selection import train_test_split
y = df['ARR_DEL15'].values
X = df.drop(['ARR_DEL15'], axis=1).values

X_train, X_test, y_train, y_test =  train_test_split(X,y, test_size=0.20, random_state=123)
In [40]:
print(X_train.shape,y_train.shape)
print(X_test.shape,y_test.shape)

(479414, 8) (479414,)
(119854, 8) (119854,)

MODEL BUILDING¶

RANDOM FOREST¶

In [41]:
from sklearn.ensemble import RandomForestClassifier

model = RandomForestClassifier(random_state=13)
model.fit(X_train, y_train)
Out[41]:
RandomForestClassifier(random_state=13)
In [42]:
RandomForestClassifier(random_state=13)
Out[42]:
RandomForestClassifier(random_state=13)
In [43]:
predicted = model.predict(X_test)
model.score(X_test, y_test)
Out[43]:
0.9125686251606121
In [46]:
#testing the model
from sklearn.metrics  import classification_report,confusion_matrix,f1_score
model_predictions = model.predict(X_test)
print("confusion matrix is:\n",confusion_matrix(y_test, model_predictions))

print("\nClassification report is:\n",classification_report(y_test, model_predictions))

confusion matrix is:
 [[101118   2222]
 [  7742   8772]]

Classification report is:
               precision    recall  f1-score   support

         0.0       0.93      0.98      0.95    103340
         1.0       0.80      0.53      0.64     16514

    accuracy                           0.92    119854
   macro avg       0.86      0.75      0.80    119854
weighted avg       0.91      0.92      0.91    119854

DECISION TREE¶

In [45]:
from sklearn.tree import DecisionTreeClassifier

model = DecisionTreeClassifier()
model.fit(X_train, y_train)
Out[45]:
DecisionTreeClassifier()
In [47]:
predicted = model.predict(X_test)
model.score(X_test, y_test)
Out[47]:
0.916865519715654
In [48]:
#testing the model
from sklearn.metrics  import classification_report,confusion_matrix,f1_score
model_predictions = model.predict(X_test)
print("confusion matrix is:\n",confusion_matrix(y_test, model_predictions))

print("\nClassification report is:\n",classification_report(y_test, model_predictions))

confusion matrix is:
 [[101118   2222]
 [  7742   8772]]

Classification report is:
               precision    recall  f1-score   support

         0.0       0.93      0.98      0.95    103340
         1.0       0.80      0.53      0.64     16514

    accuracy                           0.92    119854
   macro avg       0.86      0.75      0.80    119854
weighted avg       0.91      0.92      0.91    119854

KNEIGHBORS¶

In [49]:
from sklearn.neighbors import KNeighborsClassifier

model = KNeighborsClassifier()
model.fit(X_train, y_train)
Out[49]:
KNeighborsClassifier()
In [50]:
predicted = model.predict(X_test)
model.score(X_test, y_test)
Out[50]:
0.8730789126770905

LOGISTIC REGRESSION¶

In [52]:
from sklearn.linear_model import LogisticRegression

model = LogisticRegression()
model.fit(X_train, y_train)
Out[52]:
LogisticRegression()
In [53]:
predicted = model.predict(X_test)
model.score(X_test, y_test)
Out[53]:
0.8622156957631786
In [ ]: